Hierarchical Testing with Rabbit Optimization for Industrial Cyber-Physical Systems

TL;DR

The paper tackles the robustness of deep-learning-based Prognostics and Health Management (PHM) in Industrial Cyber-Physical Systems (ICPS) by exposing models to adversarial disturbances. It introduces HERO, a hierarchical, distribution-aware testing framework that combines Global Distribution Approximation via LSTMVAE and KDE with a Local Robustness Indicator based on gradient norms, and uses Artificial Rabbit Optimization (ARO) to generate physically constrained time-series adversarial examples. Key contributions include the four-tier HERO framework, a neuro-symbolic approach to approximate global data distributions, a gradient-based local robustness metric, physically constrained test-case generation, and empirical validation on the IEEE PHM 2014 PEMFC dataset, where HERO uncovers vulnerabilities in state-of-the-art models and demonstrates improved efficiency over prior methods. The framework is generalizable across diverse ICPS domains and offers a practical path toward more resilient PHM systems for sustainable energy applications.

Abstract

This paper presents HERO (Hierarchical Testing with Rabbit Optimization), a novel black-box adversarial testing framework for evaluating the robustness of deep learning-based Prognostics and Health Management systems in Industrial Cyber-Physical Systems. Leveraging Artificial Rabbit Optimization, HERO generates physically constrained adversarial examples that align with real-world data distributions via global and local perspective. Its generalizability ensures applicability across diverse ICPS scenarios. This study specifically focuses on the Proton Exchange Membrane Fuel Cell system, chosen for its highly dynamic operational conditions, complex degradation mechanisms, and increasing integration into ICPS as a sustainable and efficient energy solution. Experimental results highlight HERO's ability to uncover vulnerabilities in even state-of-the-art PHM models, underscoring the critical need for enhanced robustness in real-world applications. By addressing these challenges, HERO demonstrates its potential to advance more resilient PHM systems across a wide range of ICPS domains.

Figures (8)

• Figure 1: Schematic Overview of the TSTransformer Architecture.
• Figure 2: Framework of Hierarchical Testing with Rabbit Optimization
• Figure 3: Visualisation of original data, condensed data and filtered data.
• Figure 4: Feature comparsion between FC1 and FC2
• Figure 5: Similarity comparison between the adversarial example and test seed.